Research Of Feature Extraction Based On Space Handwriting Recognition

On-line handwriting recognition has been one of the most important topics in the field of pattern recognition for many years. Compared with traditional plane handwriting technology, the user experience of space handwriting is more natural, it is the trend and hot topic of handwriting recognition in recent years.Feature extraction is a key step in pattern recognition. Similarly feature extraction is also the most important step in space handwriting recognition. Whether feature is good or bad will greatly affect the performance of space handwriting recognition systems, so it is necessary to do deeply research into feature extraction methods when we research the space handwriting recognition. Some recognition methods based on time-domain feature and frequency-domain feature have been proposed. How to concurrent use time-domain feature and frequency-domain feature for space handwriting recognition systems in order to improve recognition rate is a problem to be worth studying.In order to solve the above problems, the research in this thesis is listed as follows:1. The thesis reviews the development background, application field and state-of-art of space handwriting recognition. Then the thesis introduces the entire space handwriting recognition system in details and summarizes related processing technologies in various steps of space handwriting recognition.2. The mainstream feature extraction methods for space handwriting recognition are summarized and introduced in this thesis, including time-domain features and frequency-domain features. Time-domain features are listed as follows acceleration related features, such as acceleration, velocity and position, short-time energy (STE); frequency-domain features are listed as follows:Discrete cosine transform feature(DCT),Fast Fourier transform feature(FFT) and a new hybrid feature which combines wavelet packet decomposition with Fast Fourier transform (WPD+FFT).The performance of these feature extraction methods have been compared by experiments in space handwriting character recognition and we will get the best features that can be applied in space handwriting character recognition.3. In order to make use of time-domain feature and frequency-domain feature at the same time, a feature fusion method is proposed and the principal component analysis (PCA) is employed to reduce the dimension of the fusion feature. The experimental results show that the fusion of the above two categories features can significantly improve the recognition performance of space handwriting systems because of their good complementarity.